Deskripsi Pekerjaan
Shape the Future of Generative AI
At Nexus Intelligence Labs, we don't just build models; we architect intelligence systems that redefine human-computer interaction. We are seeking a Senior Machine Learning Engineer to lead the lifecycle of our flagship Large Language Model (LLM) platform, serving millions of API requests daily.
This is a high-autonomy, high-impact role reporting directly to the VP of Engineering. You will bridge the gap between cutting-edge research (transformer architecture, RLHF, RAG optimization) and production-grade, low-latency inference infrastructure. If you have a track record of taking models from arXiv to production at scale, we want to talk.
Why This Role is Different
- Compute Access: Dedicated H100/A100 clusters; no queueing for GPU time.
- Open Culture: Publish papers, open-source tooling, speak at NeurIPS/ICML.
- Ownership: End-to-end ownership of model performance, cost optimization, and latency SLAs.
Tanggung Jawab
- Architect & Train: Design and implement distributed training pipelines (FSDP/DeepSpeed) for 7B-70B parameter models on multi-node GPU clusters.
- Optimize Inference: Deploy and optimize models using TensorRT-LLM, vLLM, and custom CUDA kernels to achieve sub-50ms TTFT (Time To First Token) at scale.
- Data Flywheel: Build automated data curation, synthetic data generation, and preference alignment (DPO/PPO) pipelines to continuously improve model quality.
- RAG & Agentic Systems: Develop advanced Retrieval-Augmented Generation pipelines with hybrid search, reranking, and agentic tool-use capabilities.
- MLOps & Observability: Establish rigorous evaluation frameworks (benchmarks, human eval, LLM-as-a-judge), model versioning, and drift detection monitoring.
- Cross-Functional Leadership: Collaborate with Product and Research to define technical roadmaps; mentor junior engineers and conduct architectural reviews.
Kualifikasi
- Education: MS/PhD in Computer Science, Machine Learning, or equivalent practical experience (5+ years industry ML experience).
- Framework Mastery: Expert-level proficiency in PyTorch/JAX; deep understanding of transformer internals, attention mechanics, and quantization (GPTQ, AWQ, GGUF).
- Distributed Systems: Proven experience with large-scale distributed training (Megatron-LM, FSDP, ZeRO) and high-throughput inference serving (vLLM, TGI, Triton).
- Production Track Record: Successfully shipped Generative AI/LLM products to production; expertise in latency optimization, GPU memory management, and cost reduction.
- MLOps Fluency: Strong hands-on experience with Kubernetes (EKS/GKE), Docker, MLflow/W&B, and feature stores.
- Software Engineering: Clean, maintainable Python code (type hints, testing, CI/CD); Rust or C++ experience for kernel work is a significant plus.